馃攳
Modelling and Forecasting using GARCH Model: Case Study on Malaysia Rubber Price - YouTube
Channel: unknown
[0]
Hi, assalamualaikum and good morning
[2]
My name is Ahmad Imran Haziq Bin Ahmad Sabri
[5]
with matric number
[7]
S52074
[9]
my supervisor is Dr. Hanafi Bin A Rahim
[12]
In this video, I want to explain about my thesis project
[17]
my thesis project title is Modelling and Forecasting using GARCH Model
[21]
Case Study on Malaysia Rubber Price
[26]
Firstly, for the introduction.
[28]
This research is to test and investigate
[30]
the Modelling and Forecasting using GARCH Model of Malaysia Rubber Price
[35]
The data for this research was taken from official portal of Malaysia
[39]
Rubber Board (MRB)
[40]
The two models
[42]
used are Generalized Autoregressive Conditional Heteroscedasticity
[45]
(GARCH) model and GARCH model with Student鈥檚 t distribution(GARCH.t)
[51]
Autocorrelation Function test (ACF),
[53]
Ljung-Box test (LB) and Lagrange Multiplier (LM) test
[56]
are used to test the existence of ARIMA effect and ARCH effect the data.
[62]
So, Akaike鈥檚 Information Criteria (AIC) and Bayesian information criteria (BIC)
[66]
are used for comparing the GARCH Model or GARCH.t
[71]
which one will perform better.
[74]
Next we will proceed with research objectives.
[77]
First, to model rubber price using GARCH model.
[80]
Second, to model rubber price using GARCH model with Student鈥檚 饾憽 distribution.
[86]
Third, to compare the GARCH model and GARCH models with Student鈥檚 饾憽 distribution.
[92]
Lastly, to forecast the volatility price of rubber for 100 day
[97]
Next, literature review for this research.
[99]
Md Ghani and Rahim (2018)
[102]
used three models ARMA, ARCH and GARCH models to study
[105]
rubber prices to predict the return fluctuations of rubber prices in the future market.
[110]
The results show that the ARMA(1,0) and GARCH(1,2) models are the best observed models.
[116]
Chin-Lin et al. (2010) investigate about the conditional correlation modeling of Asian
[121]
rubber spot and futures return volatility.
[124]
They use the VARMA-GARCH model and the VARMA-AGARCH model
[129]
to calculate rubber spot and futures returns.
[134]
The results show that the statistically significant asymmetric effect
[138]
of negative and positive shocks of the same magnitude on the conditional
[143]
variance indicates that VARMA-AGARCH is superior to its VARMA-GARCH counterpart.
[151]
Method that have been used in this study.
[154]
Firstly, transform financial series to return series use this formula.
[160]
Second method is pre-modelling test.
[163]
Autocorrelation on return data
[165]
To detect ARIMA effect on return data using Ljung-Box test
[168]
The p-value must less than 0.05
[171]
Autocorrelation on squared return data
[173]
To detect ARCH/GARCH effect on return data using Ljung-Box test
[177]
The p-value must less than 0.05
[180]
Test ARCH effect on return data
[182]
Using Langrage Multiplier. The p-value must less than 0.05
[187]
Third method is model diagnostic, Autocorrelation on error
[190]
To detect ARIMA effect on error using Ljung-Box test
[193]
Autocorrelation on squared error
[195]
To detect ARCH/GARCH effect on error using Ljung-Box test
[198]
Test ARCH effect on error , Using Langrage Multiplier
[202]
To get fit model, all p-value for the test must less than 0.05
[208]
Last method is model selection. the model will be selected
[213]
using Akaike鈥檚 Information Criteria (AIC) and Bayesian Information Criteria (BIC).
[218]
The AIC and BIC must have the smallest value to get best model.
[224]
next result & discussion, TIBSCO S+ PROGRAMMING are use to get result.
[230]
Result for pre-modelling test,
[232]
p-value for test for autocorrelation on return data is 0.00 < 0.05.
[239]
Then it is necessary to create ARIMA model.
[242]
P-value for autocorrelation on squared return data is 0.00 < 0.05,
[248]
there is ARCH effect in the return data and need to create a GARCH model.
[252]
P-value for ARCH effect on return data is 0.00<0.05,
[258]
there is ARCH effect and it necessary to create GARCH model.
[264]
For model diagnostic, p-value for autocorrelation on error is 0.00<0.05,
[271]
then the resulting model is still in need of improvement ARIMA,
[276]
but it doesn't have to be done because we only focus on garch.
[282]
P-value for autocorrelation on squared error is 0.1118>0.05,
[289]
there is no autocorrelation in the data, then the resulting model is adequate.
[294]
P-value for arch effect on error is 0.1612>0.05,
[301]
there is no autocorrelation in the data, then the resulting model is adequate.
[307]
Next, the figure show result for ARIMA(1,0,1) model.
[313]
next the figure show result for GARCH(1,1) model.
[319]
next the figure show result for GARCH(1,1) with student鈥檚 t distribution(GARCH.t).
[328]
Next, model selection. this is show value for AIC and BIC.
[334]
Referring to AIC and BIC, the estimation showed that GARCH.饾憽 is better than GARCH.
[341]
This graph show forecasting in 100 days for GARCH.
[346]
This graph show forecasting in 100 days for GARCH with student鈥檚 t distribution.
[353]
For conclusion, This research uses the framework of GARCH model and GARCH.t model
[358]
to model rubber prices based on historical information.
[362]
The modeling work was carried out in two different comparisons.
[366]
Among the GARCH model and GARCH.t model performance,
[370]
GARCH.t model shows better performance in AIC and BIC measurements.
[375]
The result is based on the lowest value performed in the AIC and BIC measurement.
[383]
That all, Thank You
Most Recent Videos:
You can go back to the homepage right here: Homepage